Instructions to use HelpingAI/Cipher-20B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use HelpingAI/Cipher-20B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="HelpingAI/Cipher-20B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("HelpingAI/Cipher-20B")
model = AutoModelForCausalLM.from_pretrained("HelpingAI/Cipher-20B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use HelpingAI/Cipher-20B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "HelpingAI/Cipher-20B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "HelpingAI/Cipher-20B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/HelpingAI/Cipher-20B

SGLang

How to use HelpingAI/Cipher-20B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "HelpingAI/Cipher-20B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "HelpingAI/Cipher-20B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "HelpingAI/Cipher-20B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "HelpingAI/Cipher-20B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use HelpingAI/Cipher-20B with Docker Model Runner:
```
docker model run hf.co/HelpingAI/Cipher-20B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

💻 Cipher-20B

[📜 License](https://helpingai.co/license) | [🌐 Website](https://helpingai.co)

🌟 About Cipher-20B

Cipher-20B is a 20 billion parameter causal language model designed for code generation.

💻 Implementation

Using Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load Cipher-20B
model = AutoModelForCausalLM.from_pretrained("HelpingAI/Cipher-20B")
tokenizer = AutoTokenizer.from_pretrained("HelpingAI/Cipher-20B")

# Example usage
code_task = [
    {"role": "system", "content": "You are Cipher"},
    {"role": "user", "content": "Write a Python function to calculate the Fibonacci sequence."}
]

inputs = tokenizer.apply_chat_template(
    code_task,
    add_generation_prompt=True,
    return_tensors="pt"
)

outputs = model.generate(
    inputs,
    max_new_tokens=256,
    temperature=0.7,
    top_p=0.9,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

⚙️ Training Details

Training Data

Trained on a large dataset of code, programming tasks, and technical documentation.
Fine-tuned for multiple programming languages like Python, JavaScript, and C++.

Capabilities

Generates code in multiple languages.
Detects and corrects common coding errors.
Provides clear explanations of code.

⚠️ Limitations

May generate verbose code depending on the input.
Long code generation may exceed token limits.
Ambiguous instructions can lead to incomplete or incorrect code.
Prioritizes efficiency in code generation.

Safety

Avoids generating harmful or malicious code.
Will not assist with illegal or unethical activities.

📚 Citation

@misc{cipher2024,
  author = {Abhay Koul},
  title = {Cipher-20B: Your Ultimate Code Buddy},
  year = {2024},
  publisher = {HelpingAI},
  journal = {HuggingFace},
  howpublished = {\url{https://huggingface.co/HelpingAI/Cipher-20B}}
}

Built with dedication, precision, and passion by HelpingAI

Website • GitHub • Discord • HuggingFace

Downloads last month: 16

Safetensors

Model size

21B params

Tensor type

BF16

Model tree for HelpingAI/Cipher-20B

Quantizations

2 models